Apple to upgrade its AI cloud computers with the M4 chip next year

Apple has developed special computers to process certain Apple Intelligence requests privately in the cloud, and these computers currently run on the M2 Ultra chip. However, it seems that the company already has plans to upgrade its cloud computers with the M4 chip starting next year.

A Nikkei Asia report on Wednesday revealed that Apple is in talks with Foxconn to build new artificial intelligence servers in Taiwan, as the company seeks to speed up the rollout of its AI-based features. Sources familiar with the matter say that Apple chose Taiwan to “tap the engineering talent and R&D resources” that work for Nvidia, which is also a Foxconn customer.

For those unfamiliar, Apple Intelligence relies on both on-device and online processing. When local language models are unable to handle a request, Apple uses its Private Cloud Compute (PCC) modules to process the request with end-to-end encryption. Here’s how Apple describes PCC:

9to5Mac first found references to Apple PCCs in early 2023. While most PCCs are equipped with the M2 Ultra chip, Apple also has some modules running on the M1 chip – most likely to handle lighter tasks. Interestingly, according to Nikkei Asia, future PCC modules to be built next year will be equipped with the latest M4 chip introduced this year rather than the M2 Ultra.

Since the M4 family of chips brings significant advances when it comes to AI tasks, it’s no surprise that Apple wants to upgrade the modules used to process data from Apple Intelligence online.

The report doesn’t detail which variant of the M4 chip Apple will use in its future PCC modules, but presumably it will be the yet-to-be-introduced M4 Ultra chip. The company recently announced the M4 Pro and M4 Max variants after announcing the M4 earlier this year.

Apple’s Security Research blog has more details on how Private Cloud Compute works, and it’s definitely worth a read.